How to Know If a Function Is Even or Odd: Symmetry Rules That Actually Work

A function is even if f(−x) = f(x) for every x in its domain, and odd if f(−x) = −f(x) for every x. Testing is mechanical: substitute −x, simplify, and compare the result with the original f(x). Even functions mirror across the y-axis; odd functions rotate onto themselves around the origin. If neither equation holds, the function has no parity. Those three sentences answer the most common version of the question you might bring to this page.

The substitution mechanism and why signs propagate the way they do

The test works because a negative input flows through every operation inside the function. An even power of a negative number is positive, so (−x)² = x² and (−x)⁴ = x⁴. An odd power keeps the minus sign, so (−x)³ = −x³. That single fact carries most of the answer. A polynomial made only of even powers is even; a polynomial made only of odd powers is odd.

Watch what happens with each operation type. Multiplication by a constant preserves parity, because a constant has no x to flip. Addition is where parity breaks: adding two even functions yields an even function, and adding two odd functions yields an odd function, but mixing the two produces neither. Composition behaves in a memorable pattern. If g is even, then f(g(x)) is even no matter what f is, because the inner function erases the sign before the outer function sees it. Composition with an odd inner function copies the outer function's parity instead.

Subtraction deserves its own sentence. Subtracting an even function from an even function keeps the result even, and the same holds for odd minus odd. Inside parentheses, x² − 4 is even and x³ − x is odd, even though each piece looks like it involves subtraction notice. The subtraction sign does not decide parity; the powers do.

Graphically the test has a physical meaning. Even symmetry means folding the plane along the y-axis maps the curve onto itself: the point (3, 9) sits alongside (−3, 9). Odd symmetry means rotating the graph 180 degrees about the origin reproduces it: the point (3, 27) pairs with (−3, −27). A mixed function such as x² + x is neither. It fails both equations for almost every x, and one counterexample is enough to end the test.

Quantified context: how rarely the symmetry equations hold

Parity is a strict condition, not a tendency. The equation f(−x) = f(x) must hold at every point of the domain; a single failure disqualifies the function. Consider x² + x. At x = 2, f(2) = 6 and f(−2) = 2, so the function fails the even test and also fails the odd test, since 2 ≠ −6. One point in a domain of infinitely many settled the verdict.

Polynomials show the same scarcity from a different angle. A polynomial of degree 5 has six coefficients, and parity requires five of the six to vanish, leaving exactly two surviving candidates: x⁵ and x³ and so on. For a polynomial of degree n, exactly one of the two parity families fits the monomials present, or none does. Roughly half of randomly assembled polynomials land in the neither category, because a single stray term of the wrong parity poisons the whole expression.

Domains impose an additional count. A function can only be even or odd on a domain symmetric about zero. The interval (−5, 5) qualifies; (−5, 5] does not, and neither does [0, ∞). Functions with restricted domains such as √x are undefined at negative inputs, so the parity question cannot even be posed for them until the domain is extended or trimmed to a symmetric set.

Edge cases: pieces, splits, and the difference quotient

Piecewise functions need each piece checked and each boundary respected. The absolute value function is the cleanest example: |x| equals x for nonnegative inputs and −x for negative inputs, and substituting −x gives |−x| = |x|, so the function is even. Its graph is a V whose two arms are mirror images.

The counterintuitive detail involves the derivative. The derivative of an even function is odd, and the derivative of an odd function is even, provided the derivative exists. Check it: x² is even, and its derivative 2x is odd. The derivative of sin x is cos x, and sine is odd while cosine is even. This pairing explains why odd functions that cross the origin must have even slopes there such as the flat tangent at the bottom of a parabola.

Odd functions carry a mandatory point at the origin. If zero belongs to the domain and f(−0) = −f(0), then f(0) must equal −f(0), which forces f(0) = 0. A function like 1/x is technically odd on its domain, since f(−x) = −f(x) wherever both sides are defined, but the domain excludes zero and the graph never touches the origin. Neither even nor odd functions can be constant unless that constant is zero in the odd case.

The difference quotient offers a shortcut and a warning. Compute f(−x) − f(x). If the result is zero for all x, the function is even. If it equals −2f(x), the function is odd. Any other outcome means neither parity applies. This form is faster for complicated expressions than comparing two full simplifications.

Misconceptions that survive most reviews

Several false rules circulate because they work on the textbook examples and fail everywhere else. Here are the ones worth flagging:

"Positive exponents mean even, negative exponents mean odd." Wrong. Parity depends on whether the exponent's effect flips the sign, not on the exponent's sign. The function x^(−2) = 1/x² is even, because squaring erases the sign, while x^(−1) = 1/x is odd. "All polynomials are even or odd." False. x² + x is a polynomial and has no parity. Most polynomials are neither. "Odd functions must pass through the origin." Only when zero is in the domain. The function 1/x is odd and leaves a gap at x = 0. "Even functions are always nonnegative." False. −x² is even and negative everywhere except the origin. "Graphs that look symmetric are symmetric." False. A curve can look mirror-symmetric at the scale of a printed graph and fail the equation far from the window. Trust the algebra.

A professional caveat closes the list. Parity is preserved by real constants but destroyed by complex ones. On the real line, multiplying by i twists an even function into an odd one, because i(−x) = −ix. If you are working numerically, check the equation at a spread of sample points including zero, one positive value, and one negative value of large magnitude, because floating-point roundoff can make a truly even function appear slightly asymmetric at extreme scales.

Practical actions: the three-step test and a comparison

Run this procedure on any candidate function:

Write the domain and confirm it is symmetric about zero. If not, the function cannot be even or odd; stop. Substitute −x everywhere x appears, then simplify fully, distributing signs and reducing powers. Compare the simplified f(−x) against f(x) and against −f(x). Match against the first means even, the second means odd, and anything else means neither.

The comparison table below collects the common families against their parity:

A partial list of common errors would include dropping the minus sign inside parentheses, forgetting that constants multiply through without changing parity, and stopping simplified comparisons at a single x value. Each error produces a confident wrong answer, which is why the second step must be finished fully before comparing.

Limitations, thresholds, and where parity ends

Parity is binary and it is fragile. One off-parity term breaks it entirely, so a function like x⁴ − x² + x is neither even nor odd despite five sixths of its monomials behaving. This is not a defect of the test; it is a consequence of the definition demanding the equation at every point.

The method also refuses to answer the symmetry question for domains that are not symmetric. If a function is defined only on [0, 6], you can still ask about parity after extending it to a symmetric interval, but not before. Functions defined on a closed interval symmetric about zero such as [−4, 4] can be tested normally.

Numerical thresholds matter in applied settings. The equation f(−x) = f(x) holds to machine precision for true even functions, but floating-point evaluation of large exponentials or small trigonometric values can show discrepancies near 10⁻¹⁵ of the input's relative scale. A professional treats a discrepancy near the square root of machine epsilon as numerical noise rather than a parity failure, but a discrepancy at the 1% level is real and disqualifies the function.

Finally, parity is a property of the function alone, not of its graph window or its tabulated samples. Two different functions can look symmetric in a narrow view and diverge far from it. The algebra, run to completion on the full domain, is the only reliable test.

For more on symmetry principles beyond parity, consult authoritative references such as the mathematical resources maintained by major scientific societies.

Limitations of this guide and how to extend the method

This article covers standard real-valued functions of one variable. It does not address parity of multivariable functions, where "even in x" and "even in y" are separate properties, or parity of sequences, where the index replaces x. It also omits the Fourier decomposition used in engineering, where every function splits into an even part and an odd part, f(x) = (f(x) + f(−x))/2 plus (f(x) − f(−x))/2, regardless of whether either part matches the original function.

To apply the method beyond this page, repeat the substitution test on each variable while holding the others fixed, and check the symmetries separately. The equations never change; only the number of inputs does.